AWS bill line item
USE1-Claude4.6Sonnet-cache-read-input-token-count
Tokens that Bedrock serves from its prompt cache appear as their own usage type. Bedrock does not reprocess them as full input. AWS bills them at a steep discount against a normal input token.
Bedrock prompt-cache read tokens
Verified against official AWS documentation last checked
Why it shows up
This line item appears as soon as an application uses prompt caching on a supported model. Teams that did not know caching was on often see a new, unfamiliar usage type on the bill. It represents savings, not new spend.
The savings come only if the cache gets hits. A bursty, low-repeat workload has more cache-write volume than cache-read volume. For that workload, the cache overhead can cost more than it saves.
How to cut it
- Check the cache read-to-write ratio before assuming caching is saving money.
- Put static, reusable content (system prompt, shared context) first in the prompt so the cache prefix matches across calls.
- Only enable caching for workloads with repeated prompt prefixes, not one-off requests.
How to read this code
USE1 / USE2us-east-1 (N. Virginia) / us-east-2 (Ohio)
Source
Source: AWS docs →Found USE1-Claude4.6Sonnet-cache-read-input-token-count on your own bill?
Drop your Cost Explorer CSV into the AWS Bill Analyzer. It'll flag line items like this one on your own bill and link straight back to the fix.
Analyze your own bill →Token spend moves every day.
Model prices and rate limits change between invoices. CostMon tracks your AI spend daily, so a change shows up before the bill does.